dir.c 40.3 KB
Newer Older
1 2
/*
  FUSE: Filesystem in Userspace
M
Miklos Szeredi 已提交
3
  Copyright (C) 2001-2008  Miklos Szeredi <miklos@szeredi.hu>
4 5 6 7 8 9 10 11 12 13 14

  This program can be distributed under the terms of the GNU GPL.
  See the file COPYING.
*/

#include "fuse_i.h"

#include <linux/pagemap.h>
#include <linux/file.h>
#include <linux/sched.h>
#include <linux/namei.h>
15
#include <linux/slab.h>
S
Seth Forshee 已提交
16
#include <linux/xattr.h>
17
#include <linux/iversion.h>
S
Seth Forshee 已提交
18
#include <linux/posix_acl.h>
19

20 21 22 23 24 25 26
static void fuse_advise_use_readdirplus(struct inode *dir)
{
	struct fuse_inode *fi = get_fuse_inode(dir);

	set_bit(FUSE_I_ADVISE_RDPLUS, &fi->state);
}

M
Miklos Szeredi 已提交
27 28 29 30 31
union fuse_dentry {
	u64 time;
	struct rcu_head rcu;
};

M
Miklos Szeredi 已提交
32 33
static inline void fuse_dentry_settime(struct dentry *entry, u64 time)
{
M
Miklos Szeredi 已提交
34
	((union fuse_dentry *) entry->d_fsdata)->time = time;
M
Miklos Szeredi 已提交
35 36 37 38
}

static inline u64 fuse_dentry_time(struct dentry *entry)
{
M
Miklos Szeredi 已提交
39
	return ((union fuse_dentry *) entry->d_fsdata)->time;
M
Miklos Szeredi 已提交
40 41
}

42 43 44
/*
 * FUSE caches dentries and attributes with separate timeout.  The
 * time in jiffies until the dentry/attributes are valid is stored in
M
Miklos Szeredi 已提交
45
 * dentry->d_fsdata and fuse_inode->i_time respectively.
46 47 48 49 50
 */

/*
 * Calculate the time in jiffies until a dentry/attributes are valid
 */
M
Miklos Szeredi 已提交
51
static u64 time_to_jiffies(u64 sec, u32 nsec)
52
{
M
Miklos Szeredi 已提交
53
	if (sec || nsec) {
M
Miklos Szeredi 已提交
54 55
		struct timespec64 ts = {
			sec,
56
			min_t(u32, nsec, NSEC_PER_SEC - 1)
M
Miklos Szeredi 已提交
57 58 59
		};

		return get_jiffies_64() + timespec64_to_jiffies(&ts);
M
Miklos Szeredi 已提交
60
	} else
M
Miklos Szeredi 已提交
61
		return 0;
62 63
}

64 65 66 67
/*
 * Set dentry and possibly attribute timeouts from the lookup/mk*
 * replies
 */
M
Miklos Szeredi 已提交
68
void fuse_change_entry_timeout(struct dentry *entry, struct fuse_entry_out *o)
69
{
M
Miklos Szeredi 已提交
70 71
	fuse_dentry_settime(entry,
		time_to_jiffies(o->entry_valid, o->entry_valid_nsec));
72 73 74 75 76 77 78
}

static u64 attr_timeout(struct fuse_attr_out *o)
{
	return time_to_jiffies(o->attr_valid, o->attr_valid_nsec);
}

M
Miklos Szeredi 已提交
79
u64 entry_attr_timeout(struct fuse_entry_out *o)
80 81
{
	return time_to_jiffies(o->attr_valid, o->attr_valid_nsec);
82 83
}

84 85 86 87
/*
 * Mark the attributes as stale, so that at the next call to
 * ->getattr() they will be fetched from userspace
 */
88 89
void fuse_invalidate_attr(struct inode *inode)
{
M
Miklos Szeredi 已提交
90
	get_fuse_inode(inode)->i_time = 0;
91 92
}

93 94 95 96 97 98
static void fuse_dir_changed(struct inode *dir)
{
	fuse_invalidate_attr(dir);
	inode_maybe_inc_iversion(dir, false);
}

99 100 101 102 103 104 105 106 107 108
/**
 * Mark the attributes as stale due to an atime change.  Avoid the invalidate if
 * atime is not used.
 */
void fuse_invalidate_atime(struct inode *inode)
{
	if (!IS_RDONLY(inode))
		fuse_invalidate_attr(inode);
}

109 110 111 112 113 114 115 116
/*
 * Just mark the entry as stale, so that a next attempt to look it up
 * will result in a new lookup call to userspace
 *
 * This is called when a dentry is about to become negative and the
 * timeout is unknown (unlink, rmdir, rename and in some cases
 * lookup)
 */
M
Miklos Szeredi 已提交
117
void fuse_invalidate_entry_cache(struct dentry *entry)
118
{
M
Miklos Szeredi 已提交
119
	fuse_dentry_settime(entry, 0);
120 121
}

122 123 124 125
/*
 * Same as fuse_invalidate_entry_cache(), but also try to remove the
 * dentry from the hash
 */
126 127 128 129
static void fuse_invalidate_entry(struct dentry *entry)
{
	d_invalidate(entry);
	fuse_invalidate_entry_cache(entry);
130 131
}

132
static void fuse_lookup_init(struct fuse_conn *fc, struct fuse_args *args,
A
Al Viro 已提交
133
			     u64 nodeid, const struct qstr *name,
134 135
			     struct fuse_entry_out *outarg)
{
136
	memset(outarg, 0, sizeof(struct fuse_entry_out));
137 138 139 140 141 142
	args->in.h.opcode = FUSE_LOOKUP;
	args->in.h.nodeid = nodeid;
	args->in.numargs = 1;
	args->in.args[0].size = name->len + 1;
	args->in.args[0].value = name->name;
	args->out.numargs = 1;
143
	args->out.args[0].size = sizeof(struct fuse_entry_out);
144
	args->out.args[0].value = outarg;
145 146
}

147
u64 fuse_get_attr_version(struct fuse_conn *fc)
148 149 150 151 152 153 154 155 156 157 158 159 160 161
{
	u64 curr_version;

	/*
	 * The spin lock isn't actually needed on 64bit archs, but we
	 * don't yet care too much about such optimizations.
	 */
	spin_lock(&fc->lock);
	curr_version = fc->attr_version;
	spin_unlock(&fc->lock);

	return curr_version;
}

162 163 164 165 166 167 168 169 170
/*
 * Check whether the dentry is still valid
 *
 * If the entry validity timeout has expired and the dentry is
 * positive, try to redo the lookup.  If the lookup results in a
 * different inode, then let the VFS invalidate the dentry and redo
 * the lookup once more.  If the lookup results in the same inode,
 * then refresh the attributes, timeouts and mark the dentry valid.
 */
171
static int fuse_dentry_revalidate(struct dentry *entry, unsigned int flags)
172
{
173
	struct inode *inode;
174 175
	struct dentry *parent;
	struct fuse_conn *fc;
M
Miklos Szeredi 已提交
176
	struct fuse_inode *fi;
177
	int ret;
178

179
	inode = d_inode_rcu(entry);
180
	if (inode && is_bad_inode(inode))
181
		goto invalid;
182 183
	else if (time_before64(fuse_dentry_time(entry), get_jiffies_64()) ||
		 (flags & LOOKUP_REVAL)) {
184
		struct fuse_entry_out outarg;
185
		FUSE_ARGS(args);
186
		struct fuse_forget_link *forget;
187
		u64 attr_version;
188

189
		/* For negative dentries, always do a fresh lookup */
190
		if (!inode)
191
			goto invalid;
192

193
		ret = -ECHILD;
194
		if (flags & LOOKUP_RCU)
195
			goto out;
196

197
		fc = get_fuse_conn(inode);
198

199
		forget = fuse_alloc_forget();
200 201
		ret = -ENOMEM;
		if (!forget)
202
			goto out;
M
Miklos Szeredi 已提交
203

204
		attr_version = fuse_get_attr_version(fc);
205

206
		parent = dget_parent(entry);
207
		fuse_lookup_init(fc, &args, get_node_id(d_inode(parent)),
208
				 &entry->d_name, &outarg);
209
		ret = fuse_simple_request(fc, &args);
210
		dput(parent);
211
		/* Zero nodeid is same as -ENOENT */
212 213 214
		if (!ret && !outarg.nodeid)
			ret = -ENOENT;
		if (!ret) {
M
Miklos Szeredi 已提交
215
			fi = get_fuse_inode(inode);
216
			if (outarg.nodeid != get_node_id(inode)) {
217
				fuse_queue_forget(fc, forget, outarg.nodeid, 1);
218
				goto invalid;
219
			}
220
			spin_lock(&fc->lock);
M
Miklos Szeredi 已提交
221
			fi->nlookup++;
222
			spin_unlock(&fc->lock);
223
		}
224
		kfree(forget);
225 226 227
		if (ret == -ENOMEM)
			goto out;
		if (ret || (outarg.attr.mode ^ inode->i_mode) & S_IFMT)
228
			goto invalid;
229

S
Seth Forshee 已提交
230
		forget_all_cached_acls(inode);
231 232 233 234
		fuse_change_attributes(inode, &outarg.attr,
				       entry_attr_timeout(&outarg),
				       attr_version);
		fuse_change_entry_timeout(entry, &outarg);
235
	} else if (inode) {
M
Miklos Szeredi 已提交
236 237 238 239 240
		fi = get_fuse_inode(inode);
		if (flags & LOOKUP_RCU) {
			if (test_bit(FUSE_I_INIT_RDPLUS, &fi->state))
				return -ECHILD;
		} else if (test_and_clear_bit(FUSE_I_INIT_RDPLUS, &fi->state)) {
241
			parent = dget_parent(entry);
242
			fuse_advise_use_readdirplus(d_inode(parent));
243 244
			dput(parent);
		}
245
	}
246 247 248 249 250 251 252
	ret = 1;
out:
	return ret;

invalid:
	ret = 0;
	goto out;
253 254
}

M
Miklos Szeredi 已提交
255 256 257 258 259 260 261 262 263 264 265 266 267
static int fuse_dentry_init(struct dentry *dentry)
{
	dentry->d_fsdata = kzalloc(sizeof(union fuse_dentry), GFP_KERNEL);

	return dentry->d_fsdata ? 0 : -ENOMEM;
}
static void fuse_dentry_release(struct dentry *dentry)
{
	union fuse_dentry *fd = dentry->d_fsdata;

	kfree_rcu(fd, rcu);
}

A
Al Viro 已提交
268
const struct dentry_operations fuse_dentry_operations = {
269
	.d_revalidate	= fuse_dentry_revalidate,
M
Miklos Szeredi 已提交
270 271
	.d_init		= fuse_dentry_init,
	.d_release	= fuse_dentry_release,
272 273
};

274 275 276 277 278
const struct dentry_operations fuse_root_dentry_operations = {
	.d_init		= fuse_dentry_init,
	.d_release	= fuse_dentry_release,
};

279
int fuse_valid_type(int m)
280 281 282 283 284
{
	return S_ISREG(m) || S_ISDIR(m) || S_ISLNK(m) || S_ISCHR(m) ||
		S_ISBLK(m) || S_ISFIFO(m) || S_ISSOCK(m);
}

A
Al Viro 已提交
285
int fuse_lookup_name(struct super_block *sb, u64 nodeid, const struct qstr *name,
286
		     struct fuse_entry_out *outarg, struct inode **inode)
287
{
288
	struct fuse_conn *fc = get_fuse_conn_super(sb);
289
	FUSE_ARGS(args);
290
	struct fuse_forget_link *forget;
291
	u64 attr_version;
292
	int err;
293

294 295 296 297
	*inode = NULL;
	err = -ENAMETOOLONG;
	if (name->len > FUSE_NAME_MAX)
		goto out;
298 299


300 301
	forget = fuse_alloc_forget();
	err = -ENOMEM;
302
	if (!forget)
303
		goto out;
M
Miklos Szeredi 已提交
304

305
	attr_version = fuse_get_attr_version(fc);
306

307 308
	fuse_lookup_init(fc, &args, nodeid, name, outarg);
	err = fuse_simple_request(fc, &args);
309
	/* Zero nodeid is same as -ENOENT, but with valid timeout */
310 311 312 313 314 315 316 317 318 319 320 321 322 323
	if (err || !outarg->nodeid)
		goto out_put_forget;

	err = -EIO;
	if (!outarg->nodeid)
		goto out_put_forget;
	if (!fuse_valid_type(outarg->attr.mode))
		goto out_put_forget;

	*inode = fuse_iget(sb, outarg->nodeid, outarg->generation,
			   &outarg->attr, entry_attr_timeout(outarg),
			   attr_version);
	err = -ENOMEM;
	if (!*inode) {
324
		fuse_queue_forget(fc, forget, outarg->nodeid, 1);
325
		goto out;
326
	}
327 328 329
	err = 0;

 out_put_forget:
330
	kfree(forget);
331 332 333 334 335
 out:
	return err;
}

static struct dentry *fuse_lookup(struct inode *dir, struct dentry *entry,
A
Al Viro 已提交
336
				  unsigned int flags)
337 338 339 340 341 342
{
	int err;
	struct fuse_entry_out outarg;
	struct inode *inode;
	struct dentry *newent;
	bool outarg_valid = true;
343
	bool locked;
344

345
	locked = fuse_lock_inode(dir);
346 347
	err = fuse_lookup_name(dir->i_sb, get_node_id(dir), &entry->d_name,
			       &outarg, &inode);
348
	fuse_unlock_inode(dir, locked);
349 350 351 352 353 354 355 356 357 358
	if (err == -ENOENT) {
		outarg_valid = false;
		err = 0;
	}
	if (err)
		goto out_err;

	err = -EIO;
	if (inode && get_node_id(inode) == FUSE_ROOT_ID)
		goto out_iput;
359

360
	newent = d_splice_alias(inode, entry);
361 362 363
	err = PTR_ERR(newent);
	if (IS_ERR(newent))
		goto out_err;
364

365
	entry = newent ? newent : entry;
366
	if (outarg_valid)
367
		fuse_change_entry_timeout(entry, &outarg);
368 369
	else
		fuse_invalidate_entry_cache(entry);
370

371
	fuse_advise_use_readdirplus(dir);
372
	return newent;
373 374 375 376 377

 out_iput:
	iput(inode);
 out_err:
	return ERR_PTR(err);
378 379
}

380 381 382 383 384 385
/*
 * Atomic create+open operation
 *
 * If the filesystem doesn't support this, then fall back to separate
 * 'mknod' + 'open' requests.
 */
A
Al Viro 已提交
386
static int fuse_create_open(struct inode *dir, struct dentry *entry,
A
Al Viro 已提交
387
			    struct file *file, unsigned flags,
388
			    umode_t mode)
M
Miklos Szeredi 已提交
389 390 391 392
{
	int err;
	struct inode *inode;
	struct fuse_conn *fc = get_fuse_conn(dir);
393
	FUSE_ARGS(args);
394
	struct fuse_forget_link *forget;
395
	struct fuse_create_in inarg;
M
Miklos Szeredi 已提交
396 397 398 399
	struct fuse_open_out outopen;
	struct fuse_entry_out outentry;
	struct fuse_file *ff;

400 401 402
	/* Userspace expects S_IFREG in create mode */
	BUG_ON((mode & S_IFMT) != S_IFREG);

403
	forget = fuse_alloc_forget();
404
	err = -ENOMEM;
405
	if (!forget)
406
		goto out_err;
407

408
	err = -ENOMEM;
T
Tejun Heo 已提交
409
	ff = fuse_file_alloc(fc);
M
Miklos Szeredi 已提交
410
	if (!ff)
411
		goto out_put_forget_req;
M
Miklos Szeredi 已提交
412

413 414 415
	if (!fc->dont_mask)
		mode &= ~current_umask();

M
Miklos Szeredi 已提交
416 417
	flags &= ~O_NOCTTY;
	memset(&inarg, 0, sizeof(inarg));
418
	memset(&outentry, 0, sizeof(outentry));
M
Miklos Szeredi 已提交
419 420
	inarg.flags = flags;
	inarg.mode = mode;
421
	inarg.umask = current_umask();
422 423 424
	args.in.h.opcode = FUSE_CREATE;
	args.in.h.nodeid = get_node_id(dir);
	args.in.numargs = 2;
425
	args.in.args[0].size = sizeof(inarg);
426 427 428 429
	args.in.args[0].value = &inarg;
	args.in.args[1].size = entry->d_name.len + 1;
	args.in.args[1].value = entry->d_name.name;
	args.out.numargs = 2;
430
	args.out.args[0].size = sizeof(outentry);
431 432 433 434
	args.out.args[0].value = &outentry;
	args.out.args[1].size = sizeof(outopen);
	args.out.args[1].value = &outopen;
	err = fuse_simple_request(fc, &args);
435
	if (err)
M
Miklos Szeredi 已提交
436 437 438
		goto out_free_ff;

	err = -EIO;
439
	if (!S_ISREG(outentry.attr.mode) || invalid_nodeid(outentry.nodeid))
M
Miklos Szeredi 已提交
440 441
		goto out_free_ff;

442 443 444
	ff->fh = outopen.fh;
	ff->nodeid = outentry.nodeid;
	ff->open_flags = outopen.open_flags;
M
Miklos Szeredi 已提交
445
	inode = fuse_iget(dir->i_sb, outentry.nodeid, outentry.generation,
446
			  &outentry.attr, entry_attr_timeout(&outentry), 0);
M
Miklos Szeredi 已提交
447 448
	if (!inode) {
		flags &= ~(O_CREAT | O_EXCL | O_TRUNC);
449
		fuse_sync_release(ff, flags);
450
		fuse_queue_forget(fc, forget, outentry.nodeid, 1);
451 452
		err = -ENOMEM;
		goto out_err;
M
Miklos Szeredi 已提交
453
	}
454
	kfree(forget);
M
Miklos Szeredi 已提交
455
	d_instantiate(entry, inode);
456
	fuse_change_entry_timeout(entry, &outentry);
457
	fuse_dir_changed(dir);
458
	err = finish_open(file, entry, generic_file_open);
A
Al Viro 已提交
459
	if (err) {
460
		fuse_sync_release(ff, flags);
461
	} else {
462
		file->private_data = ff;
463
		fuse_finish_open(inode, file);
M
Miklos Szeredi 已提交
464
	}
A
Al Viro 已提交
465
	return err;
M
Miklos Szeredi 已提交
466

467
out_free_ff:
M
Miklos Szeredi 已提交
468
	fuse_file_free(ff);
469
out_put_forget_req:
470
	kfree(forget);
471
out_err:
A
Al Viro 已提交
472
	return err;
473 474 475
}

static int fuse_mknod(struct inode *, struct dentry *, umode_t, dev_t);
A
Al Viro 已提交
476
static int fuse_atomic_open(struct inode *dir, struct dentry *entry,
A
Al Viro 已提交
477
			    struct file *file, unsigned flags,
478
			    umode_t mode)
479 480 481 482 483
{
	int err;
	struct fuse_conn *fc = get_fuse_conn(dir);
	struct dentry *res = NULL;

484
	if (d_in_lookup(entry)) {
A
Al Viro 已提交
485
		res = fuse_lookup(dir, entry, 0);
486
		if (IS_ERR(res))
A
Al Viro 已提交
487
			return PTR_ERR(res);
488 489 490 491 492

		if (res)
			entry = res;
	}

493
	if (!(flags & O_CREAT) || d_really_is_positive(entry))
494 495 496
		goto no_open;

	/* Only creates */
497
	file->f_mode |= FMODE_CREATED;
498 499 500 501

	if (fc->no_create)
		goto mknod;

502
	err = fuse_create_open(dir, entry, file, flags, mode);
A
Al Viro 已提交
503
	if (err == -ENOSYS) {
504 505 506 507 508
		fc->no_create = 1;
		goto mknod;
	}
out_dput:
	dput(res);
A
Al Viro 已提交
509
	return err;
510 511 512

mknod:
	err = fuse_mknod(dir, entry, mode, 0);
A
Al Viro 已提交
513
	if (err)
514 515
		goto out_dput;
no_open:
A
Al Viro 已提交
516
	return finish_no_open(file, res);
M
Miklos Szeredi 已提交
517 518
}

519 520 521
/*
 * Code shared between mknod, mkdir, symlink and link
 */
522
static int create_new_entry(struct fuse_conn *fc, struct fuse_args *args,
523
			    struct inode *dir, struct dentry *entry,
A
Al Viro 已提交
524
			    umode_t mode)
525 526 527
{
	struct fuse_entry_out outarg;
	struct inode *inode;
A
Al Viro 已提交
528
	struct dentry *d;
529
	int err;
530
	struct fuse_forget_link *forget;
M
Miklos Szeredi 已提交
531

532
	forget = fuse_alloc_forget();
533
	if (!forget)
534
		return -ENOMEM;
535

536
	memset(&outarg, 0, sizeof(outarg));
537 538
	args->in.h.nodeid = get_node_id(dir);
	args->out.numargs = 1;
539
	args->out.args[0].size = sizeof(outarg);
540 541
	args->out.args[0].value = &outarg;
	err = fuse_simple_request(fc, args);
M
Miklos Szeredi 已提交
542 543 544
	if (err)
		goto out_put_forget_req;

545 546
	err = -EIO;
	if (invalid_nodeid(outarg.nodeid))
M
Miklos Szeredi 已提交
547
		goto out_put_forget_req;
548 549

	if ((outarg.attr.mode ^ mode) & S_IFMT)
M
Miklos Szeredi 已提交
550
		goto out_put_forget_req;
551

552
	inode = fuse_iget(dir->i_sb, outarg.nodeid, outarg.generation,
553
			  &outarg.attr, entry_attr_timeout(&outarg), 0);
554
	if (!inode) {
555
		fuse_queue_forget(fc, forget, outarg.nodeid, 1);
556 557
		return -ENOMEM;
	}
558
	kfree(forget);
559

A
Al Viro 已提交
560 561 562 563
	d_drop(entry);
	d = d_splice_alias(inode, entry);
	if (IS_ERR(d))
		return PTR_ERR(d);
564

A
Al Viro 已提交
565 566 567 568 569 570
	if (d) {
		fuse_change_entry_timeout(d, &outarg);
		dput(d);
	} else {
		fuse_change_entry_timeout(entry, &outarg);
	}
571
	fuse_dir_changed(dir);
572
	return 0;
573

M
Miklos Szeredi 已提交
574
 out_put_forget_req:
575
	kfree(forget);
576
	return err;
577 578
}

A
Al Viro 已提交
579
static int fuse_mknod(struct inode *dir, struct dentry *entry, umode_t mode,
580 581 582 583
		      dev_t rdev)
{
	struct fuse_mknod_in inarg;
	struct fuse_conn *fc = get_fuse_conn(dir);
584
	FUSE_ARGS(args);
585

586 587 588
	if (!fc->dont_mask)
		mode &= ~current_umask();

589 590 591
	memset(&inarg, 0, sizeof(inarg));
	inarg.mode = mode;
	inarg.rdev = new_encode_dev(rdev);
592
	inarg.umask = current_umask();
593 594
	args.in.h.opcode = FUSE_MKNOD;
	args.in.numargs = 2;
595
	args.in.args[0].size = sizeof(inarg);
596 597 598 599
	args.in.args[0].value = &inarg;
	args.in.args[1].size = entry->d_name.len + 1;
	args.in.args[1].value = entry->d_name.name;
	return create_new_entry(fc, &args, dir, entry, mode);
600 601
}

A
Al Viro 已提交
602
static int fuse_create(struct inode *dir, struct dentry *entry, umode_t mode,
A
Al Viro 已提交
603
		       bool excl)
604 605 606 607
{
	return fuse_mknod(dir, entry, mode, 0);
}

608
static int fuse_mkdir(struct inode *dir, struct dentry *entry, umode_t mode)
609 610 611
{
	struct fuse_mkdir_in inarg;
	struct fuse_conn *fc = get_fuse_conn(dir);
612
	FUSE_ARGS(args);
613

614 615 616
	if (!fc->dont_mask)
		mode &= ~current_umask();

617 618
	memset(&inarg, 0, sizeof(inarg));
	inarg.mode = mode;
619
	inarg.umask = current_umask();
620 621 622 623 624 625 626
	args.in.h.opcode = FUSE_MKDIR;
	args.in.numargs = 2;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	args.in.args[1].size = entry->d_name.len + 1;
	args.in.args[1].value = entry->d_name.name;
	return create_new_entry(fc, &args, dir, entry, S_IFDIR);
627 628 629 630 631 632 633
}

static int fuse_symlink(struct inode *dir, struct dentry *entry,
			const char *link)
{
	struct fuse_conn *fc = get_fuse_conn(dir);
	unsigned len = strlen(link) + 1;
634
	FUSE_ARGS(args);
635

636 637 638 639 640 641 642
	args.in.h.opcode = FUSE_SYMLINK;
	args.in.numargs = 2;
	args.in.args[0].size = entry->d_name.len + 1;
	args.in.args[0].value = entry->d_name.name;
	args.in.args[1].size = len;
	args.in.args[1].value = link;
	return create_new_entry(fc, &args, dir, entry, S_IFLNK);
643 644
}

S
Seth Forshee 已提交
645
void fuse_update_ctime(struct inode *inode)
M
Maxim Patlasov 已提交
646 647
{
	if (!IS_NOCMTIME(inode)) {
648
		inode->i_ctime = current_time(inode);
M
Maxim Patlasov 已提交
649 650 651 652
		mark_inode_dirty_sync(inode);
	}
}

653 654 655 656
static int fuse_unlink(struct inode *dir, struct dentry *entry)
{
	int err;
	struct fuse_conn *fc = get_fuse_conn(dir);
657 658 659 660 661 662 663 664
	FUSE_ARGS(args);

	args.in.h.opcode = FUSE_UNLINK;
	args.in.h.nodeid = get_node_id(dir);
	args.in.numargs = 1;
	args.in.args[0].size = entry->d_name.len + 1;
	args.in.args[0].value = entry->d_name.name;
	err = fuse_simple_request(fc, &args);
665
	if (!err) {
666
		struct inode *inode = d_inode(entry);
M
Miklos Szeredi 已提交
667
		struct fuse_inode *fi = get_fuse_inode(inode);
668

M
Miklos Szeredi 已提交
669 670
		spin_lock(&fc->lock);
		fi->attr_version = ++fc->attr_version;
671 672 673 674 675 676 677 678
		/*
		 * If i_nlink == 0 then unlink doesn't make sense, yet this can
		 * happen if userspace filesystem is careless.  It would be
		 * difficult to enforce correct nlink usage so just ignore this
		 * condition here
		 */
		if (inode->i_nlink > 0)
			drop_nlink(inode);
M
Miklos Szeredi 已提交
679
		spin_unlock(&fc->lock);
680
		fuse_invalidate_attr(inode);
681
		fuse_dir_changed(dir);
682
		fuse_invalidate_entry_cache(entry);
M
Maxim Patlasov 已提交
683
		fuse_update_ctime(inode);
684 685 686 687 688 689 690 691 692
	} else if (err == -EINTR)
		fuse_invalidate_entry(entry);
	return err;
}

static int fuse_rmdir(struct inode *dir, struct dentry *entry)
{
	int err;
	struct fuse_conn *fc = get_fuse_conn(dir);
693 694 695 696 697 698 699 700
	FUSE_ARGS(args);

	args.in.h.opcode = FUSE_RMDIR;
	args.in.h.nodeid = get_node_id(dir);
	args.in.numargs = 1;
	args.in.args[0].size = entry->d_name.len + 1;
	args.in.args[0].value = entry->d_name.name;
	err = fuse_simple_request(fc, &args);
701
	if (!err) {
702
		clear_nlink(d_inode(entry));
703
		fuse_dir_changed(dir);
704
		fuse_invalidate_entry_cache(entry);
705 706 707 708 709
	} else if (err == -EINTR)
		fuse_invalidate_entry(entry);
	return err;
}

M
Miklos Szeredi 已提交
710 711 712
static int fuse_rename_common(struct inode *olddir, struct dentry *oldent,
			      struct inode *newdir, struct dentry *newent,
			      unsigned int flags, int opcode, size_t argsize)
713 714
{
	int err;
M
Miklos Szeredi 已提交
715
	struct fuse_rename2_in inarg;
716
	struct fuse_conn *fc = get_fuse_conn(olddir);
717
	FUSE_ARGS(args);
718

M
Miklos Szeredi 已提交
719
	memset(&inarg, 0, argsize);
720
	inarg.newdir = get_node_id(newdir);
M
Miklos Szeredi 已提交
721
	inarg.flags = flags;
722 723 724 725 726 727 728 729 730 731
	args.in.h.opcode = opcode;
	args.in.h.nodeid = get_node_id(olddir);
	args.in.numargs = 3;
	args.in.args[0].size = argsize;
	args.in.args[0].value = &inarg;
	args.in.args[1].size = oldent->d_name.len + 1;
	args.in.args[1].value = oldent->d_name.name;
	args.in.args[2].size = newent->d_name.len + 1;
	args.in.args[2].value = newent->d_name.name;
	err = fuse_simple_request(fc, &args);
732
	if (!err) {
733
		/* ctime changes */
734 735
		fuse_invalidate_attr(d_inode(oldent));
		fuse_update_ctime(d_inode(oldent));
736

M
Miklos Szeredi 已提交
737
		if (flags & RENAME_EXCHANGE) {
738 739
			fuse_invalidate_attr(d_inode(newent));
			fuse_update_ctime(d_inode(newent));
M
Miklos Szeredi 已提交
740 741
		}

742
		fuse_dir_changed(olddir);
743
		if (olddir != newdir)
744
			fuse_dir_changed(newdir);
745 746

		/* newent will end up negative */
747 748
		if (!(flags & RENAME_EXCHANGE) && d_really_is_positive(newent)) {
			fuse_invalidate_attr(d_inode(newent));
749
			fuse_invalidate_entry_cache(newent);
750
			fuse_update_ctime(d_inode(newent));
M
Miklos Szeredi 已提交
751
		}
752 753 754 755 756 757 758
	} else if (err == -EINTR) {
		/* If request was interrupted, DEITY only knows if the
		   rename actually took place.  If the invalidation
		   fails (e.g. some process has CWD under the renamed
		   directory), then there can be inconsistency between
		   the dcache and the real filesystem.  Tough luck. */
		fuse_invalidate_entry(oldent);
759
		if (d_really_is_positive(newent))
760 761 762 763 764 765
			fuse_invalidate_entry(newent);
	}

	return err;
}

M
Miklos Szeredi 已提交
766 767 768 769 770 771 772 773 774 775
static int fuse_rename2(struct inode *olddir, struct dentry *oldent,
			struct inode *newdir, struct dentry *newent,
			unsigned int flags)
{
	struct fuse_conn *fc = get_fuse_conn(olddir);
	int err;

	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
		return -EINVAL;

M
Miklos Szeredi 已提交
776 777 778
	if (flags) {
		if (fc->no_rename2 || fc->minor < 23)
			return -EINVAL;
M
Miklos Szeredi 已提交
779

M
Miklos Szeredi 已提交
780 781 782 783 784 785 786 787 788 789 790
		err = fuse_rename_common(olddir, oldent, newdir, newent, flags,
					 FUSE_RENAME2,
					 sizeof(struct fuse_rename2_in));
		if (err == -ENOSYS) {
			fc->no_rename2 = 1;
			err = -EINVAL;
		}
	} else {
		err = fuse_rename_common(olddir, oldent, newdir, newent, 0,
					 FUSE_RENAME,
					 sizeof(struct fuse_rename_in));
M
Miklos Szeredi 已提交
791
	}
M
Miklos Szeredi 已提交
792

M
Miklos Szeredi 已提交
793
	return err;
M
Miklos Szeredi 已提交
794
}
M
Miklos Szeredi 已提交
795

796 797 798 799 800
static int fuse_link(struct dentry *entry, struct inode *newdir,
		     struct dentry *newent)
{
	int err;
	struct fuse_link_in inarg;
801
	struct inode *inode = d_inode(entry);
802
	struct fuse_conn *fc = get_fuse_conn(inode);
803
	FUSE_ARGS(args);
804 805 806

	memset(&inarg, 0, sizeof(inarg));
	inarg.oldnodeid = get_node_id(inode);
807 808 809 810 811 812 813
	args.in.h.opcode = FUSE_LINK;
	args.in.numargs = 2;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	args.in.args[1].size = newent->d_name.len + 1;
	args.in.args[1].value = newent->d_name.name;
	err = create_new_entry(fc, &args, newdir, newent, inode->i_mode);
814 815 816 817 818 819
	/* Contrary to "normal" filesystems it can happen that link
	   makes two "logical" inodes point to the same "physical"
	   inode.  We invalidate the attributes of the old one, so it
	   will reflect changes in the backing inode (link count,
	   etc.)
	*/
M
Miklos Szeredi 已提交
820 821 822 823 824 825 826
	if (!err) {
		struct fuse_inode *fi = get_fuse_inode(inode);

		spin_lock(&fc->lock);
		fi->attr_version = ++fc->attr_version;
		inc_nlink(inode);
		spin_unlock(&fc->lock);
827
		fuse_invalidate_attr(inode);
M
Maxim Patlasov 已提交
828
		fuse_update_ctime(inode);
M
Miklos Szeredi 已提交
829 830 831
	} else if (err == -EINTR) {
		fuse_invalidate_attr(inode);
	}
832 833 834
	return err;
}

835 836 837
static void fuse_fillattr(struct inode *inode, struct fuse_attr *attr,
			  struct kstat *stat)
{
M
Miklos Szeredi 已提交
838
	unsigned int blkbits;
P
Pavel Emelyanov 已提交
839 840 841
	struct fuse_conn *fc = get_fuse_conn(inode);

	/* see the comment in fuse_change_attributes() */
M
Maxim Patlasov 已提交
842
	if (fc->writeback_cache && S_ISREG(inode->i_mode)) {
P
Pavel Emelyanov 已提交
843
		attr->size = i_size_read(inode);
M
Maxim Patlasov 已提交
844 845
		attr->mtime = inode->i_mtime.tv_sec;
		attr->mtimensec = inode->i_mtime.tv_nsec;
M
Maxim Patlasov 已提交
846 847
		attr->ctime = inode->i_ctime.tv_sec;
		attr->ctimensec = inode->i_ctime.tv_nsec;
M
Maxim Patlasov 已提交
848
	}
M
Miklos Szeredi 已提交
849

850 851 852 853
	stat->dev = inode->i_sb->s_dev;
	stat->ino = attr->ino;
	stat->mode = (inode->i_mode & S_IFMT) | (attr->mode & 07777);
	stat->nlink = attr->nlink;
854 855
	stat->uid = make_kuid(fc->user_ns, attr->uid);
	stat->gid = make_kgid(fc->user_ns, attr->gid);
856 857 858 859 860 861 862 863 864
	stat->rdev = inode->i_rdev;
	stat->atime.tv_sec = attr->atime;
	stat->atime.tv_nsec = attr->atimensec;
	stat->mtime.tv_sec = attr->mtime;
	stat->mtime.tv_nsec = attr->mtimensec;
	stat->ctime.tv_sec = attr->ctime;
	stat->ctime.tv_nsec = attr->ctimensec;
	stat->size = attr->size;
	stat->blocks = attr->blocks;
M
Miklos Szeredi 已提交
865 866 867 868 869 870 871

	if (attr->blksize != 0)
		blkbits = ilog2(attr->blksize);
	else
		blkbits = inode->i_sb->s_blocksize_bits;

	stat->blksize = 1 << blkbits;
872 873
}

874 875
static int fuse_do_getattr(struct inode *inode, struct kstat *stat,
			   struct file *file)
876 877
{
	int err;
878 879
	struct fuse_getattr_in inarg;
	struct fuse_attr_out outarg;
880
	struct fuse_conn *fc = get_fuse_conn(inode);
881
	FUSE_ARGS(args);
882 883
	u64 attr_version;

884
	attr_version = fuse_get_attr_version(fc);
885

886
	memset(&inarg, 0, sizeof(inarg));
887
	memset(&outarg, 0, sizeof(outarg));
888 889 890 891 892 893 894
	/* Directories have separate file-handle space */
	if (file && S_ISREG(inode->i_mode)) {
		struct fuse_file *ff = file->private_data;

		inarg.getattr_flags |= FUSE_GETATTR_FH;
		inarg.fh = ff->fh;
	}
895 896 897 898 899 900
	args.in.h.opcode = FUSE_GETATTR;
	args.in.h.nodeid = get_node_id(inode);
	args.in.numargs = 1;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	args.out.numargs = 1;
901
	args.out.args[0].size = sizeof(outarg);
902 903
	args.out.args[0].value = &outarg;
	err = fuse_simple_request(fc, &args);
904
	if (!err) {
905
		if ((inode->i_mode ^ outarg.attr.mode) & S_IFMT) {
906 907 908
			make_bad_inode(inode);
			err = -EIO;
		} else {
909 910
			fuse_change_attributes(inode, &outarg.attr,
					       attr_timeout(&outarg),
911 912
					       attr_version);
			if (stat)
913
				fuse_fillattr(inode, &outarg.attr, stat);
914 915 916 917 918
		}
	}
	return err;
}

M
Miklos Szeredi 已提交
919
static int fuse_update_get_attr(struct inode *inode, struct file *file,
M
Miklos Szeredi 已提交
920
				struct kstat *stat, unsigned int flags)
M
Miklos Szeredi 已提交
921 922
{
	struct fuse_inode *fi = get_fuse_inode(inode);
M
Miklos Szeredi 已提交
923
	int err = 0;
M
Miklos Szeredi 已提交
924
	bool sync;
M
Miklos Szeredi 已提交
925

M
Miklos Szeredi 已提交
926 927 928 929 930 931 932 933
	if (flags & AT_STATX_FORCE_SYNC)
		sync = true;
	else if (flags & AT_STATX_DONT_SYNC)
		sync = false;
	else
		sync = time_before64(fi->i_time, get_jiffies_64());

	if (sync) {
S
Seth Forshee 已提交
934
		forget_all_cached_acls(inode);
M
Miklos Szeredi 已提交
935
		err = fuse_do_getattr(inode, stat, file);
M
Miklos Szeredi 已提交
936 937 938 939
	} else if (stat) {
		generic_fillattr(inode, stat);
		stat->mode = fi->orig_i_mode;
		stat->ino = fi->orig_ino;
M
Miklos Szeredi 已提交
940 941 942 943 944
	}

	return err;
}

M
Miklos Szeredi 已提交
945 946
int fuse_update_attributes(struct inode *inode, struct file *file)
{
M
Miklos Szeredi 已提交
947
	return fuse_update_get_attr(inode, file, NULL, 0);
M
Miklos Szeredi 已提交
948 949
}

J
John Muir 已提交
950
int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
951
			     u64 child_nodeid, struct qstr *name)
J
John Muir 已提交
952 953 954 955 956 957 958 959 960 961
{
	int err = -ENOTDIR;
	struct inode *parent;
	struct dentry *dir;
	struct dentry *entry;

	parent = ilookup5(sb, parent_nodeid, fuse_inode_eq, &parent_nodeid);
	if (!parent)
		return -ENOENT;

A
Al Viro 已提交
962
	inode_lock(parent);
J
John Muir 已提交
963 964 965 966 967 968 969 970
	if (!S_ISDIR(parent->i_mode))
		goto unlock;

	err = -ENOENT;
	dir = d_find_alias(parent);
	if (!dir)
		goto unlock;

971
	name->hash = full_name_hash(dir, name->name, name->len);
J
John Muir 已提交
972 973 974 975 976
	entry = d_lookup(dir, name);
	dput(dir);
	if (!entry)
		goto unlock;

977
	fuse_dir_changed(parent);
J
John Muir 已提交
978
	fuse_invalidate_entry(entry);
979

980
	if (child_nodeid != 0 && d_really_is_positive(entry)) {
A
Al Viro 已提交
981
		inode_lock(d_inode(entry));
982
		if (get_node_id(d_inode(entry)) != child_nodeid) {
983 984 985 986 987 988 989
			err = -ENOENT;
			goto badentry;
		}
		if (d_mountpoint(entry)) {
			err = -EBUSY;
			goto badentry;
		}
990
		if (d_is_dir(entry)) {
991 992 993 994 995
			shrink_dcache_parent(entry);
			if (!simple_empty(entry)) {
				err = -ENOTEMPTY;
				goto badentry;
			}
996
			d_inode(entry)->i_flags |= S_DEAD;
997 998
		}
		dont_mount(entry);
999
		clear_nlink(d_inode(entry));
1000 1001
		err = 0;
 badentry:
A
Al Viro 已提交
1002
		inode_unlock(d_inode(entry));
1003 1004 1005 1006 1007
		if (!err)
			d_delete(entry);
	} else {
		err = 0;
	}
J
John Muir 已提交
1008 1009 1010
	dput(entry);

 unlock:
A
Al Viro 已提交
1011
	inode_unlock(parent);
J
John Muir 已提交
1012 1013 1014 1015
	iput(parent);
	return err;
}

1016 1017
/*
 * Calling into a user-controlled filesystem gives the filesystem
1018
 * daemon ptrace-like capabilities over the current process.  This
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
 * means, that the filesystem daemon is able to record the exact
 * filesystem operations performed, and can also control the behavior
 * of the requester process in otherwise impossible ways.  For example
 * it can delay the operation for arbitrary length of time allowing
 * DoS against the requester.
 *
 * For this reason only those processes can call into the filesystem,
 * for which the owner of the mount has ptrace privilege.  This
 * excludes processes started by other users, suid or sgid processes.
 */
1029
int fuse_allow_current_process(struct fuse_conn *fc)
1030
{
1031
	const struct cred *cred;
1032

M
Miklos Szeredi 已提交
1033
	if (fc->allow_other)
1034
		return current_in_userns(fc->user_ns);
1035

1036
	cred = current_cred();
1037 1038 1039 1040 1041 1042
	if (uid_eq(cred->euid, fc->user_id) &&
	    uid_eq(cred->suid, fc->user_id) &&
	    uid_eq(cred->uid,  fc->user_id) &&
	    gid_eq(cred->egid, fc->group_id) &&
	    gid_eq(cred->sgid, fc->group_id) &&
	    gid_eq(cred->gid,  fc->group_id))
1043
		return 1;
1044

1045
	return 0;
1046 1047
}

M
Miklos Szeredi 已提交
1048 1049 1050
static int fuse_access(struct inode *inode, int mask)
{
	struct fuse_conn *fc = get_fuse_conn(inode);
1051
	FUSE_ARGS(args);
M
Miklos Szeredi 已提交
1052 1053 1054
	struct fuse_access_in inarg;
	int err;

1055 1056
	BUG_ON(mask & MAY_NOT_BLOCK);

M
Miklos Szeredi 已提交
1057 1058 1059 1060
	if (fc->no_access)
		return 0;

	memset(&inarg, 0, sizeof(inarg));
1061
	inarg.mask = mask & (MAY_READ | MAY_WRITE | MAY_EXEC);
1062 1063 1064 1065 1066 1067
	args.in.h.opcode = FUSE_ACCESS;
	args.in.h.nodeid = get_node_id(inode);
	args.in.numargs = 1;
	args.in.args[0].size = sizeof(inarg);
	args.in.args[0].value = &inarg;
	err = fuse_simple_request(fc, &args);
M
Miklos Szeredi 已提交
1068 1069 1070 1071 1072 1073 1074
	if (err == -ENOSYS) {
		fc->no_access = 1;
		err = 0;
	}
	return err;
}

1075
static int fuse_perm_getattr(struct inode *inode, int mask)
1076
{
1077
	if (mask & MAY_NOT_BLOCK)
1078 1079
		return -ECHILD;

S
Seth Forshee 已提交
1080
	forget_all_cached_acls(inode);
1081 1082 1083
	return fuse_do_getattr(inode, NULL, NULL);
}

1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
/*
 * Check permission.  The two basic access models of FUSE are:
 *
 * 1) Local access checking ('default_permissions' mount option) based
 * on file mode.  This is the plain old disk filesystem permission
 * modell.
 *
 * 2) "Remote" access checking, where server is responsible for
 * checking permission in each inode operation.  An exception to this
 * is if ->permission() was invoked from sys_access() in which case an
 * access request is sent.  Execute permission is still checked
 * locally based on file mode.
 */
1097
static int fuse_permission(struct inode *inode, int mask)
1098 1099
{
	struct fuse_conn *fc = get_fuse_conn(inode);
1100 1101
	bool refreshed = false;
	int err = 0;
1102

1103
	if (!fuse_allow_current_process(fc))
1104
		return -EACCES;
1105 1106

	/*
1107
	 * If attributes are needed, refresh them before proceeding
1108
	 */
M
Miklos Szeredi 已提交
1109
	if (fc->default_permissions ||
1110
	    ((mask & MAY_EXEC) && S_ISREG(inode->i_mode))) {
1111 1112
		struct fuse_inode *fi = get_fuse_inode(inode);

M
Miklos Szeredi 已提交
1113
		if (time_before64(fi->i_time, get_jiffies_64())) {
1114 1115
			refreshed = true;

1116
			err = fuse_perm_getattr(inode, mask);
1117 1118 1119
			if (err)
				return err;
		}
1120 1121
	}

M
Miklos Szeredi 已提交
1122
	if (fc->default_permissions) {
1123
		err = generic_permission(inode, mask);
M
Miklos Szeredi 已提交
1124 1125 1126 1127

		/* If permission is denied, try to refresh file
		   attributes.  This is also needed, because the root
		   node will at first have no permissions */
1128
		if (err == -EACCES && !refreshed) {
1129
			err = fuse_perm_getattr(inode, mask);
M
Miklos Szeredi 已提交
1130
			if (!err)
1131
				err = generic_permission(inode, mask);
M
Miklos Szeredi 已提交
1132 1133
		}

1134 1135 1136 1137
		/* Note: the opposite of the above test does not
		   exist.  So if permissions are revoked this won't be
		   noticed immediately, only after the attribute
		   timeout has expired */
E
Eric Paris 已提交
1138
	} else if (mask & (MAY_ACCESS | MAY_CHDIR)) {
1139 1140 1141 1142 1143 1144
		err = fuse_access(inode, mask);
	} else if ((mask & MAY_EXEC) && S_ISREG(inode->i_mode)) {
		if (!(inode->i_mode & S_IXUGO)) {
			if (refreshed)
				return -EACCES;

1145
			err = fuse_perm_getattr(inode, mask);
1146 1147 1148
			if (!err && !(inode->i_mode & S_IXUGO))
				return -EACCES;
		}
1149
	}
1150
	return err;
1151 1152
}

1153
static const char *fuse_get_link(struct dentry *dentry,
1154 1155
				 struct inode *inode,
				 struct delayed_call *done)
1156 1157
{
	struct fuse_conn *fc = get_fuse_conn(inode);
1158
	FUSE_ARGS(args);
1159
	char *link;
1160
	ssize_t ret;
1161

1162 1163 1164
	if (!dentry)
		return ERR_PTR(-ECHILD);

A
Al Viro 已提交
1165
	link = kmalloc(PAGE_SIZE, GFP_KERNEL);
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
	if (!link)
		return ERR_PTR(-ENOMEM);

	args.in.h.opcode = FUSE_READLINK;
	args.in.h.nodeid = get_node_id(inode);
	args.out.argvar = 1;
	args.out.numargs = 1;
	args.out.args[0].size = PAGE_SIZE - 1;
	args.out.args[0].value = link;
	ret = fuse_simple_request(fc, &args);
	if (ret < 0) {
A
Al Viro 已提交
1177
		kfree(link);
1178 1179 1180
		link = ERR_PTR(ret);
	} else {
		link[ret] = '\0';
1181
		set_delayed_call(done, kfree_link, link);
1182
	}
1183
	fuse_invalidate_atime(inode);
1184 1185 1186 1187 1188
	return link;
}

static int fuse_dir_open(struct inode *inode, struct file *file)
{
1189
	return fuse_open_common(inode, file, true);
1190 1191 1192 1193
}

static int fuse_dir_release(struct inode *inode, struct file *file)
{
1194 1195 1196
	fuse_release_common(file, FUSE_RELEASEDIR);

	return 0;
1197 1198
}

1199 1200
static int fuse_dir_fsync(struct file *file, loff_t start, loff_t end,
			  int datasync)
1201
{
1202
	return fuse_fsync_common(file, start, end, datasync, 1);
1203 1204
}

1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
static long fuse_dir_ioctl(struct file *file, unsigned int cmd,
			    unsigned long arg)
{
	struct fuse_conn *fc = get_fuse_conn(file->f_mapping->host);

	/* FUSE_IOCTL_DIR only supported for API version >= 7.18 */
	if (fc->minor < 18)
		return -ENOTTY;

	return fuse_ioctl_common(file, cmd, arg, FUSE_IOCTL_DIR);
}

static long fuse_dir_compat_ioctl(struct file *file, unsigned int cmd,
				   unsigned long arg)
{
	struct fuse_conn *fc = get_fuse_conn(file->f_mapping->host);

	if (fc->minor < 18)
		return -ENOTTY;

	return fuse_ioctl_common(file, cmd, arg,
				 FUSE_IOCTL_COMPAT | FUSE_IOCTL_DIR);
}

M
Maxim Patlasov 已提交
1229
static bool update_mtime(unsigned ivalid, bool trust_local_mtime)
M
Miklos Szeredi 已提交
1230 1231 1232 1233 1234
{
	/* Always update if mtime is explicitly set  */
	if (ivalid & ATTR_MTIME_SET)
		return true;

M
Maxim Patlasov 已提交
1235 1236 1237 1238
	/* Or if kernel i_mtime is the official one */
	if (trust_local_mtime)
		return true;

M
Miklos Szeredi 已提交
1239 1240 1241 1242 1243 1244 1245 1246
	/* If it's an open(O_TRUNC) or an ftruncate(), don't update */
	if ((ivalid & ATTR_SIZE) && (ivalid & (ATTR_OPEN | ATTR_FILE)))
		return false;

	/* In all other cases update */
	return true;
}

1247 1248
static void iattr_to_fattr(struct fuse_conn *fc, struct iattr *iattr,
			   struct fuse_setattr_in *arg, bool trust_local_cmtime)
1249 1250 1251 1252
{
	unsigned ivalid = iattr->ia_valid;

	if (ivalid & ATTR_MODE)
1253
		arg->valid |= FATTR_MODE,   arg->mode = iattr->ia_mode;
1254
	if (ivalid & ATTR_UID)
1255
		arg->valid |= FATTR_UID,    arg->uid = from_kuid(fc->user_ns, iattr->ia_uid);
1256
	if (ivalid & ATTR_GID)
1257
		arg->valid |= FATTR_GID,    arg->gid = from_kgid(fc->user_ns, iattr->ia_gid);
1258
	if (ivalid & ATTR_SIZE)
1259
		arg->valid |= FATTR_SIZE,   arg->size = iattr->ia_size;
M
Miklos Szeredi 已提交
1260 1261
	if (ivalid & ATTR_ATIME) {
		arg->valid |= FATTR_ATIME;
1262
		arg->atime = iattr->ia_atime.tv_sec;
M
Miklos Szeredi 已提交
1263 1264 1265 1266
		arg->atimensec = iattr->ia_atime.tv_nsec;
		if (!(ivalid & ATTR_ATIME_SET))
			arg->valid |= FATTR_ATIME_NOW;
	}
1267
	if ((ivalid & ATTR_MTIME) && update_mtime(ivalid, trust_local_cmtime)) {
M
Miklos Szeredi 已提交
1268
		arg->valid |= FATTR_MTIME;
1269
		arg->mtime = iattr->ia_mtime.tv_sec;
M
Miklos Szeredi 已提交
1270
		arg->mtimensec = iattr->ia_mtime.tv_nsec;
1271
		if (!(ivalid & ATTR_MTIME_SET) && !trust_local_cmtime)
M
Miklos Szeredi 已提交
1272
			arg->valid |= FATTR_MTIME_NOW;
1273
	}
1274 1275 1276 1277 1278
	if ((ivalid & ATTR_CTIME) && trust_local_cmtime) {
		arg->valid |= FATTR_CTIME;
		arg->ctime = iattr->ia_ctime.tv_sec;
		arg->ctimensec = iattr->ia_ctime.tv_nsec;
	}
1279 1280
}

M
Miklos Szeredi 已提交
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291
/*
 * Prevent concurrent writepages on inode
 *
 * This is done by adding a negative bias to the inode write counter
 * and waiting for all pending writes to finish.
 */
void fuse_set_nowrite(struct inode *inode)
{
	struct fuse_conn *fc = get_fuse_conn(inode);
	struct fuse_inode *fi = get_fuse_inode(inode);

A
Al Viro 已提交
1292
	BUG_ON(!inode_is_locked(inode));
M
Miklos Szeredi 已提交
1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324

	spin_lock(&fc->lock);
	BUG_ON(fi->writectr < 0);
	fi->writectr += FUSE_NOWRITE;
	spin_unlock(&fc->lock);
	wait_event(fi->page_waitq, fi->writectr == FUSE_NOWRITE);
}

/*
 * Allow writepages on inode
 *
 * Remove the bias from the writecounter and send any queued
 * writepages.
 */
static void __fuse_release_nowrite(struct inode *inode)
{
	struct fuse_inode *fi = get_fuse_inode(inode);

	BUG_ON(fi->writectr != FUSE_NOWRITE);
	fi->writectr = 0;
	fuse_flush_writepages(inode);
}

void fuse_release_nowrite(struct inode *inode)
{
	struct fuse_conn *fc = get_fuse_conn(inode);

	spin_lock(&fc->lock);
	__fuse_release_nowrite(inode);
	spin_unlock(&fc->lock);
}

1325
static void fuse_setattr_fill(struct fuse_conn *fc, struct fuse_args *args,
M
Maxim Patlasov 已提交
1326 1327 1328 1329
			      struct inode *inode,
			      struct fuse_setattr_in *inarg_p,
			      struct fuse_attr_out *outarg_p)
{
1330 1331 1332 1333 1334 1335
	args->in.h.opcode = FUSE_SETATTR;
	args->in.h.nodeid = get_node_id(inode);
	args->in.numargs = 1;
	args->in.args[0].size = sizeof(*inarg_p);
	args->in.args[0].value = inarg_p;
	args->out.numargs = 1;
1336
	args->out.args[0].size = sizeof(*outarg_p);
1337
	args->out.args[0].value = outarg_p;
M
Maxim Patlasov 已提交
1338 1339 1340 1341 1342
}

/*
 * Flush inode->i_mtime to the server
 */
1343
int fuse_flush_times(struct inode *inode, struct fuse_file *ff)
M
Maxim Patlasov 已提交
1344 1345
{
	struct fuse_conn *fc = get_fuse_conn(inode);
1346
	FUSE_ARGS(args);
M
Maxim Patlasov 已提交
1347 1348 1349 1350 1351 1352
	struct fuse_setattr_in inarg;
	struct fuse_attr_out outarg;

	memset(&inarg, 0, sizeof(inarg));
	memset(&outarg, 0, sizeof(outarg));

1353
	inarg.valid = FATTR_MTIME;
M
Maxim Patlasov 已提交
1354 1355
	inarg.mtime = inode->i_mtime.tv_sec;
	inarg.mtimensec = inode->i_mtime.tv_nsec;
1356 1357 1358 1359 1360
	if (fc->minor >= 23) {
		inarg.valid |= FATTR_CTIME;
		inarg.ctime = inode->i_ctime.tv_sec;
		inarg.ctimensec = inode->i_ctime.tv_nsec;
	}
M
Miklos Szeredi 已提交
1361 1362 1363 1364
	if (ff) {
		inarg.valid |= FATTR_FH;
		inarg.fh = ff->fh;
	}
1365
	fuse_setattr_fill(fc, &args, inode, &inarg, &outarg);
M
Maxim Patlasov 已提交
1366

1367
	return fuse_simple_request(fc, &args);
M
Maxim Patlasov 已提交
1368 1369
}

1370 1371 1372 1373 1374
/*
 * Set attributes, and at the same time refresh them.
 *
 * Truncation is slightly complicated, because the 'truncate' request
 * may fail, in which case we don't want to touch the mapping.
M
Miklos Szeredi 已提交
1375 1376
 * vmtruncate() doesn't allow for this case, so do the rlimit checking
 * and the actual truncation by hand.
1377
 */
1378
int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,
1379
		    struct file *file)
1380
{
1381
	struct inode *inode = d_inode(dentry);
1382
	struct fuse_conn *fc = get_fuse_conn(inode);
1383
	struct fuse_inode *fi = get_fuse_inode(inode);
1384
	FUSE_ARGS(args);
1385 1386
	struct fuse_setattr_in inarg;
	struct fuse_attr_out outarg;
M
Miklos Szeredi 已提交
1387
	bool is_truncate = false;
P
Pavel Emelyanov 已提交
1388
	bool is_wb = fc->writeback_cache;
M
Miklos Szeredi 已提交
1389
	loff_t oldsize;
1390
	int err;
1391
	bool trust_local_cmtime = is_wb && S_ISREG(inode->i_mode);
1392

M
Miklos Szeredi 已提交
1393
	if (!fc->default_permissions)
1394 1395
		attr->ia_valid |= ATTR_FORCE;

1396
	err = setattr_prepare(dentry, attr);
1397 1398
	if (err)
		return err;
M
Miklos Szeredi 已提交
1399

M
Miklos Szeredi 已提交
1400
	if (attr->ia_valid & ATTR_OPEN) {
1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411
		/* This is coming from open(..., ... | O_TRUNC); */
		WARN_ON(!(attr->ia_valid & ATTR_SIZE));
		WARN_ON(attr->ia_size != 0);
		if (fc->atomic_o_trunc) {
			/*
			 * No need to send request to userspace, since actual
			 * truncation has already been done by OPEN.  But still
			 * need to truncate page cache.
			 */
			i_size_write(inode, 0);
			truncate_pagecache(inode, 0);
M
Miklos Szeredi 已提交
1412
			return 0;
1413
		}
M
Miklos Szeredi 已提交
1414 1415
		file = NULL;
	}
1416

1417
	if (attr->ia_valid & ATTR_SIZE)
M
Miklos Szeredi 已提交
1418
		is_truncate = true;
1419

1420
	if (is_truncate) {
M
Miklos Szeredi 已提交
1421
		fuse_set_nowrite(inode);
1422
		set_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
1423 1424
		if (trust_local_cmtime && attr->ia_size != inode->i_size)
			attr->ia_valid |= ATTR_MTIME | ATTR_CTIME;
1425
	}
M
Miklos Szeredi 已提交
1426

1427
	memset(&inarg, 0, sizeof(inarg));
1428
	memset(&outarg, 0, sizeof(outarg));
1429
	iattr_to_fattr(fc, attr, &inarg, trust_local_cmtime);
1430 1431 1432 1433 1434
	if (file) {
		struct fuse_file *ff = file->private_data;
		inarg.valid |= FATTR_FH;
		inarg.fh = ff->fh;
	}
1435 1436 1437 1438 1439
	if (attr->ia_valid & ATTR_SIZE) {
		/* For mandatory locking in truncate */
		inarg.valid |= FATTR_LOCKOWNER;
		inarg.lock_owner = fuse_lock_owner_id(fc, current->files);
	}
1440 1441
	fuse_setattr_fill(fc, &args, inode, &inarg, &outarg);
	err = fuse_simple_request(fc, &args);
1442 1443 1444
	if (err) {
		if (err == -EINTR)
			fuse_invalidate_attr(inode);
M
Miklos Szeredi 已提交
1445
		goto error;
1446
	}
1447

1448 1449
	if ((inode->i_mode ^ outarg.attr.mode) & S_IFMT) {
		make_bad_inode(inode);
M
Miklos Szeredi 已提交
1450 1451 1452 1453 1454
		err = -EIO;
		goto error;
	}

	spin_lock(&fc->lock);
M
Maxim Patlasov 已提交
1455
	/* the kernel maintains i_mtime locally */
1456 1457 1458 1459 1460
	if (trust_local_cmtime) {
		if (attr->ia_valid & ATTR_MTIME)
			inode->i_mtime = attr->ia_mtime;
		if (attr->ia_valid & ATTR_CTIME)
			inode->i_ctime = attr->ia_ctime;
M
Miklos Szeredi 已提交
1461
		/* FIXME: clear I_DIRTY_SYNC? */
M
Maxim Patlasov 已提交
1462 1463
	}

M
Miklos Szeredi 已提交
1464 1465 1466
	fuse_change_attributes_common(inode, &outarg.attr,
				      attr_timeout(&outarg));
	oldsize = inode->i_size;
P
Pavel Emelyanov 已提交
1467 1468 1469
	/* see the comment in fuse_change_attributes() */
	if (!is_wb || is_truncate || !S_ISREG(inode->i_mode))
		i_size_write(inode, outarg.attr.size);
M
Miklos Szeredi 已提交
1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480

	if (is_truncate) {
		/* NOTE: this may release/reacquire fc->lock */
		__fuse_release_nowrite(inode);
	}
	spin_unlock(&fc->lock);

	/*
	 * Only call invalidate_inode_pages2() after removing
	 * FUSE_NOWRITE, otherwise fuse_launder_page() would deadlock.
	 */
P
Pavel Emelyanov 已提交
1481 1482
	if ((is_truncate || !is_wb) &&
	    S_ISREG(inode->i_mode) && oldsize != outarg.attr.size) {
1483
		truncate_pagecache(inode, outarg.attr.size);
M
Miklos Szeredi 已提交
1484
		invalidate_inode_pages2(inode->i_mapping);
1485 1486
	}

1487
	clear_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
1488
	return 0;
M
Miklos Szeredi 已提交
1489 1490 1491 1492 1493

error:
	if (is_truncate)
		fuse_release_nowrite(inode);

1494
	clear_bit(FUSE_I_SIZE_UNSTABLE, &fi->state);
M
Miklos Szeredi 已提交
1495
	return err;
1496 1497
}

1498 1499
static int fuse_setattr(struct dentry *entry, struct iattr *attr)
{
1500
	struct inode *inode = d_inode(entry);
1501
	struct fuse_conn *fc = get_fuse_conn(inode);
1502
	struct file *file = (attr->ia_valid & ATTR_FILE) ? attr->ia_file : NULL;
1503
	int ret;
1504 1505 1506 1507

	if (!fuse_allow_current_process(get_fuse_conn(inode)))
		return -EACCES;

1508 1509 1510
	if (attr->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID)) {
		attr->ia_valid &= ~(ATTR_KILL_SUID | ATTR_KILL_SGID |
				    ATTR_MODE);
1511

1512
		/*
1513 1514 1515 1516
		 * The only sane way to reliably kill suid/sgid is to do it in
		 * the userspace filesystem
		 *
		 * This should be done on write(), truncate() and chown().
1517
		 */
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
		if (!fc->handle_killpriv) {
			/*
			 * ia_mode calculation may have used stale i_mode.
			 * Refresh and recalculate.
			 */
			ret = fuse_do_getattr(inode, NULL, file);
			if (ret)
				return ret;

			attr->ia_mode = inode->i_mode;
1528
			if (inode->i_mode & S_ISUID) {
1529 1530 1531
				attr->ia_valid |= ATTR_MODE;
				attr->ia_mode &= ~S_ISUID;
			}
1532
			if ((inode->i_mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1533 1534 1535
				attr->ia_valid |= ATTR_MODE;
				attr->ia_mode &= ~S_ISGID;
			}
1536 1537 1538 1539
		}
	}
	if (!attr->ia_valid)
		return 0;
1540

1541
	ret = fuse_do_setattr(entry, attr, file);
1542
	if (!ret) {
S
Seth Forshee 已提交
1543 1544 1545 1546 1547 1548 1549
		/*
		 * If filesystem supports acls it may have updated acl xattrs in
		 * the filesystem, so forget cached acls for the inode.
		 */
		if (fc->posix_acl)
			forget_all_cached_acls(inode);

1550 1551 1552 1553 1554
		/* Directory mode changed, may need to revalidate access */
		if (d_is_dir(entry) && (attr->ia_valid & ATTR_MODE))
			fuse_invalidate_entry_cache(entry);
	}
	return ret;
1555 1556
}

1557 1558
static int fuse_getattr(const struct path *path, struct kstat *stat,
			u32 request_mask, unsigned int flags)
1559
{
1560
	struct inode *inode = d_inode(path->dentry);
1561 1562
	struct fuse_conn *fc = get_fuse_conn(inode);

1563
	if (!fuse_allow_current_process(fc))
1564 1565
		return -EACCES;

M
Miklos Szeredi 已提交
1566
	return fuse_update_get_attr(inode, NULL, stat, flags);
1567 1568
}

1569
static const struct inode_operations fuse_dir_inode_operations = {
1570
	.lookup		= fuse_lookup,
1571 1572 1573 1574
	.mkdir		= fuse_mkdir,
	.symlink	= fuse_symlink,
	.unlink		= fuse_unlink,
	.rmdir		= fuse_rmdir,
1575
	.rename		= fuse_rename2,
1576 1577 1578
	.link		= fuse_link,
	.setattr	= fuse_setattr,
	.create		= fuse_create,
1579
	.atomic_open	= fuse_atomic_open,
1580
	.mknod		= fuse_mknod,
1581 1582
	.permission	= fuse_permission,
	.getattr	= fuse_getattr,
1583
	.listxattr	= fuse_listxattr,
S
Seth Forshee 已提交
1584 1585
	.get_acl	= fuse_get_acl,
	.set_acl	= fuse_set_acl,
1586 1587
};

1588
static const struct file_operations fuse_dir_operations = {
M
Miklos Szeredi 已提交
1589
	.llseek		= generic_file_llseek,
1590
	.read		= generic_read_dir,
A
Al Viro 已提交
1591
	.iterate_shared	= fuse_readdir,
1592 1593
	.open		= fuse_dir_open,
	.release	= fuse_dir_release,
1594
	.fsync		= fuse_dir_fsync,
1595 1596
	.unlocked_ioctl	= fuse_dir_ioctl,
	.compat_ioctl	= fuse_dir_compat_ioctl,
1597 1598
};

1599
static const struct inode_operations fuse_common_inode_operations = {
1600
	.setattr	= fuse_setattr,
1601 1602
	.permission	= fuse_permission,
	.getattr	= fuse_getattr,
1603
	.listxattr	= fuse_listxattr,
S
Seth Forshee 已提交
1604 1605
	.get_acl	= fuse_get_acl,
	.set_acl	= fuse_set_acl,
1606 1607
};

1608
static const struct inode_operations fuse_symlink_inode_operations = {
1609
	.setattr	= fuse_setattr,
1610
	.get_link	= fuse_get_link,
1611
	.getattr	= fuse_getattr,
1612
	.listxattr	= fuse_listxattr,
1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629
};

void fuse_init_common(struct inode *inode)
{
	inode->i_op = &fuse_common_inode_operations;
}

void fuse_init_dir(struct inode *inode)
{
	inode->i_op = &fuse_dir_inode_operations;
	inode->i_fop = &fuse_dir_operations;
}

void fuse_init_symlink(struct inode *inode)
{
	inode->i_op = &fuse_symlink_inode_operations;
}